Reissuance Concept Presentation to the NCI Board of ... · Added PROSPR data on frequency of events (abnl, bx, colposcopy) $15,260/QALY –ve $19,530/QALY vs no screening benefit

Reissuance Concept Presentation to the NCI Board of Scientific Advisors

1

Stephen Taplin, MD, MPH Paul Doria-Rose, DVM, PhD

Sarah Kobrin, PhD Paul Pinsky, PhD

Division of Cancer Control and Population Sciences

and Division of Cancer Prevention

2

Mortality reductions are possible through screening Randomized trial results

Breast (mammography) 8 RCTs – 15% mortality1

40-49 RR death 0.85 (0.75-0.96)

50-74 RR death 0.78 (0.70-0.87)*

Colon (l FOBT)** – 13-21% mortality after 18 yrs 4 trials - (RR death 0.85 (0.78-0.92)

New tests (FIT) higher sensitivity & specificity

Lung (spiral CT)& – 3 trials +, 19% mortality 4 trials – (RR death 0.81 (0.72-0.91)

Population-based observation (Pap test) Cervix – 20-60% mortality

Cervix – 90% Cervix cancer

1Nelson 2009, *2002 analysis in Nelson et al 2009, ** Whitlock 2008, & Humphrey 2013

3

Screening is process

Patient Adherence

Recruitment

Effort -In reach

-Out reach

Access to Care

Test Sensitivity /specificity

Technical

application of the test

Interpretation

Biologic

characteristics

Follow-up of

Abnormal

Notification

System

Patient

Adherence

Provider

Communication

Diagnostic Evaluation

Sensitivity- specificity

Technical

Resources

Interpretation

Tissue sample error

Follow-up of Diagnostic Evaluations

Notification System

Patient

Adherence

Provider Communication

Risk

Assessment Diagnosis

Cancer or

Precursor

Treatment Detection

56%cx

52%brst

32%cx

40% brst

13% cx

8% Brst

Invasive

Cx C

Late stage

Breast

And that process breaks down

4 4

PROSPR I – 2011-2016

Document the screening process across 3 cancers • Breast, Cervical, and Colorectal (CRC) • 7 centers funded in 2011 (U54)

• 2 supplements for Cervical cancer 2013

• 1 coordinating center (U01) • Conduct projects relevant to understanding and improving the process (U54)

5 5

The centers capture large diverse populations

6

Breast Age 18-89 n=309,346

Cervical Age 18-89 n=3,169,645

Colorectal Age 50-89 N=2,381,109

US 2010 Census**

Caucasian race*

74% 50% 59% 79-81%

African-American race*

16% 8% 9% 11-12%

Hispanic ethnicity*

4% 25% 19% 8-14%

* The balance to achieve 100% includes Asian Pacific Islander, Alaska Natives and those reporting multiple races **Females age 20+ and overall population age 50+

7

Breast – 309,346 ♀ Cervical - 3,169,645 ♀ Colorectal – 2,381,109

Provider /facility

6482 / 221 31,219/2,788 23,110/ 641

Lower Upper Lower Upper Lower Upper

% abnl 8.6% 10.7% 2.4% 6.3% 4.1% 7.0%

% eval 95% 98% 57% 84% 39% 76%

% Rxed 95% 100% 89% 100% 88% 98%

7

Screening Process Variations Can Have a large Impact

Breast Cancer has the least variation

(2.4%-6.3%) * 3,169,645 = 123,616 women (76%-39%)82,381,109 = 881,010 people

8

Organ-based projects – 9 cross-center (35)

Breast – 34 Pubs/13 in progress Tomosynthesis vs screen/film – McCarthy et al JNCI 2014

Digital breast tomosynthesis (n =15,571) vs digital mammography (n =10,728)

Reduced recall (8.8% vs 10.4% p<0.001) – Penn Verified in PROSPR study (8.7% vs 10.4% p<0.0001)

Cervical – 7 Pubs/ 9 in progress

High-value improvements in US Screening Process ? Kim et al - Annals Int Med 2015 – Disease model

Added PROSPR data on frequency of events (abnl, bx, colposcopy)

$15,260/QALY –ve $19,530/QALY vs no screening benefit in Quality Adjusted Life years > for adherence to 3 yr & bx

Colorectal - 33 Pubs/ 9 in progress –

Adenoma Detection rate – NEJM 2014 314,872 colonoscopies by 136 gastroenterologists with 712 interval ca Each 1% in adenoma detection = 3% interval ca

3 8

9

Variation in F/u after abnormal screen (abnl) – Tosteson et al

9

Time (months) Ticks at 3 and 9 months

breast

Colorectal

cervical

• 7 cross-organ papers published – 14 in process

• JGIM 12/2015 • Time –to F/u after

• Abnl mammogram • Abnl Fit/FOBT • Abnl PAP

• Breast – mature • Colorectal – variation

within site (Kaisers) • Cervical – slowest, now

adding site data

10

What we still need to know What organizational and provider factors affect the screening process?

Organizational and provider team variables not recorded in automated records and not standardized

Comparison across cancers

Breast vs. CRC vs. Cervical

Lung cancer screening

False positive evaluation effect

How to measure the quality of the screening process

Effectiveness, safety, patient-centeredness, timeliness, efficiency, equity

Long-term effects of screening

Overdiagnosis?

Adverse events among people who were screened negative

What interventions can improve the screening process?

11

PROSPR Reissuance Greater emphasis on disparities

Expand data available for screening studies

o Increase longitudinal follow-up

o Add Lung cancer screening

Establish metrics of patient, provider and system factors that affect the screening process

Evaluate quality of the screening process

o Effectiveness, safety, patient-centeredness, timeliness, efficiency, equity

Intervene at some step in the process after screening occurs

11

12

Research Centers (U54) 4 research centers (one cancer type per center)

At least 2 systems of care (collaborative application)

Representation of diverse populations

$12M annual set aside

Coordinating Center (U01) Data aggregation

Annual export of dataset

Oversight of quality measurement across cancers

$1.5M annual set aside

Organization & Funding

12

13

Potential impact of PROSPR

13

After PROSPR, we should have:

Organizational and provider factors that can be changed to improve screening

Ways of addressing differential screening across race/ethnicity

Common measures of quality

Ways to intervene upon steps in the process

Ways to measure and achieve improved screening in the United States

14

14

15

Income across PROSPR centers

15

16

Geographic location & PROSPR Centers

16

17

Insurance Coverage & PROSPR Centers

17

18

PROSPR II: Data Infrastructure Research Centers

Colorectal

Breast

Cervical

Coordinating Center

Pooled Data Resource

Multilevel Data Capture During

Screening Process

Facility

Characteristics

Provider Characteristics

and functon

Procedure Performance/ Interpretation

Procedure Results

Risk Factors, Demographics, Screening Hx

UUNM

UUNM

Lung

Pu

bli

c D

atab

ase

at N

CI

Health care settings

Primary care practices

Integrated delivery systems

Imaging centers

Ambulatory surgical centers

Federally Qualified Health Centers

We know more about parts of the process

Communication

strategy

Primary

Care

Integrated

Risk

assess

tool

CE DBT vs Digital

Current practice of self-

sampling

Molec Char of

FIT-missed

C-oscopy Qual

Molecular &

Morph DCIS

Risk

Assessment Diagnosis Cancer or

Precursor

Treatment

Detection

Zapka et al 2003

Breast x x x x x x

Cervical 0 0 x x

CRC x 0 x x x

CE of current practice

Long-term CE

of risk stratified screen

Faliure

analysis

2 CER of FIT v C-oscopy

Organization & clinic-level

factors

Characterize

populations

at low risk

for CRC

Personalize

with

BCI

Breast

Density,

collagen

reorg &

survival

Ben & Harms of

personalized DCIS

Personalized risk &

next steps for

adenoma \

Effect

Population diversity differs across centers

20

Breast Cervical Colorectal

Site A

%

Site B

%

Site C

%

Site D

%

Site E

%

Site F

%

Site G

%

Site D

%

Site E

%

Site G

%

White 79 53 95 74 45 59 10 80 55 18

Black 8 36 0 5 9 2 25 4 9 37

Hispanic 9 2 2 6 29 34 61 4 22 39

Asian/PI 4 4 1 11 15 1 3 8 13 6

Am.Ind./AK

Native

/Other

0 5 2 5 1 4 0 3 1 0